Database Reference
In-Depth Information
Gutierrez:
What does a typical day look like for you at MailChimp?
Foreman:
I lead the data science team at MailChimp, and I like to get my hands
dirty too. Some of the big pieces of my day involve working with my team to take
stock of current projects and figure out where to go next, doing my own work
and prototyping things—I do projects just like my peers do—and then also my
talking with other teams, talking with management, and planning for the future.
On our team, we've got different folks facing different kinds of projects. We've got
one person who really owns compliance and looks at our compliance processes.
We've got another data scientist who focuses on more of the user experience
side of the house and understanding our customers. I help them and others as is
needed while I do the three things mentioned earlier—build data products, be a
translator, and have conversations with the data science community.
We also have developers on the team who can take prototypes and turn them
into production tools.
And on top of that, we have a group of qualitative researchers on the team who
work alongside us to source ideas and projects—and to verify findings and
product directions—through surveys and a heck of a lot of customer visits.
That's one thing that makes the MailChimp data science unique—we've placed
quantitative and qualitative research in a single place right next to some great
engineers. It's a powerful mix.
Getting back to how my day is laid out, first thing when I get into work I take
stock of where we're at across the various projects that we work on. It's
important to understand where we're at on projects—whether it's surveying
users, mining data, or building models. Once our progress is clear to me,
I carve out time to prototype before my brain gets fried. I am not a developer,
and I don't claim to be one. However, I can write terrible spaghetti code that
will run, though no one will probably want to read it.
And so one of the things I try to do is get my hands dirty and actually get into
the data and build prototype models. An example prototype I put together
recently focused on helping people segment their list based on predicted
email address demographics—for example, age range.
Once the day is going, a big component is communicating with other teams
to figure out what we can do to help them. For instance, we may be working
with the integrations and partnerships team, who might have a big user who's
interested in doing a spherical K-means clustering of their list to investigate
different segments. So we'll work together to figure out if we can run the
clustering, and then talk through the results with them to figure out what
came out of it and who the readers are on this list. This type of work then
benefits this particular team, because they can provide this as a report to the
customer. This benefits me because I learn a little bit more about the power
of our data, as well as what our customers want and need.
